ResearchBenchmarksPairwise preference prediction on CHiLi M simulated (test)Follow96.29AccuracyPrefSQA85.463688.274391.08593.8957Jun 17, 2026Evaluation ResultsMethodMethodLinksAccuracyPrefSQApretrained encoders=fi...pretrained encoders=fine-tuned2026.0696.29SQAPP2026.0694.78PrefSQA-Fpretrained encoders=fr...pretrained encoders=frozen2026.0691.52UPPSQA2026.0685.88